Professional Documents
Culture Documents
1900–2005
Roger A. Pielke Jr.1; Joel Gratz2; Christopher W. Landsea3; Douglas Collins4; Mark A. Saunders5; and
Rade Musulin6
Abstract: After more than two decades of relatively little Atlantic hurricane activity, the past decade saw heightened hurricane activity
and more than $150 billion in damage in 2004 and 2005. This paper normalizes mainland U.S. hurricane damage from 1900–2005 to 2005
values using two methodologies. A normalization provides an estimate of the damage that would occur if storms from the past made
landfall under another year’s societal conditions. Our methods use changes in inflation and wealth at the national level and changes in
population and housing units at the coastal county level. Across both normalization methods, there is no remaining trend of increasing
absolute damage in the data set, which follows the lack of trends in landfall frequency or intensity observed over the twentieth century.
The 1970s and 1980s were notable because of the extremely low amounts of damage compared to other decades. The decade 1996–2005
has the second most damage among the past 11 decades, with only the decade 1926–1935 surpassing its costs. Over the 106 years of
record, the average annual normalized damage in the continental United States is about $10 billion under both methods. The most
damaging single storm is the 1926 Great Miami storm, with $140–157 billion of normalized damage: the most damaging years are 1926
and 2005. Of the total damage, about 85% is accounted for by the intense hurricanes 共Saffir-Simpson Categories 3, 4, and 5兲, yet these
have comprised only 24% of the U.S. landfalling tropical cyclones. Unless action is taken to address the growing concentration of people
and properties in coastal areas where hurricanes strike, damage will increase, and by a great deal, as more and wealthier people
increasingly inhabit these coastal locations.
DOI: 10.1061/共ASCE兲1527-6988共2008兲9:1共29兲
CE Database subject headings: Damage; Hurricanes; United States; Economic factors; Population growth; History.
Inflation
In order to adjust for changes in national inflation 共i.e., the de-
crease in value of a currency over time兲, we use the implicit price
deflator for gross domestic product 共IPDGDP兲 for the years 1929–
2005 from the Bureau of Economic Analysis 共BEA 2006b兲. For
inflation data for 1900–1928 the BEA recommends Johnston and
Williamson 共2006兲 as there are no official government inflation
data during these years 共V. Mannering, personal communication,
January 19, 2006兲. From these data, the inflation adjustment is a
ratio of the 2005 IPDGDP to that in the year in which the storm
made landfall. For example, the 2005 IPDGDP was 112.737 and
Fig. 1. 2005 Population by county. Galveston/Houston area of Texas,
that for 1979 was 49.548. Thus, to convert damages expressed in
Tampa and Miami area of Florida, and Northeast coastline stand out
1979 dollars to 2005 dollars requires that 1979 damages be mul- as areas with high vulnerability due to exceedingly large populations.
tiplied by 2.275= 112.737/ 49.548.
multiplier is different, however, as it corrects for national changes The general formula for the CL05 normalized losses is
in housing units—rather than population—to determine a change
in wealth per housing unit. D2005 = Dy ⫻ Iy ⫻ RWPHUy ⫻ HU2005/y 共2兲
For example, wealth in 2005 was $40.99 trillion and $8.912
where D2005 = normalized damages in 2005 dollars; Dy = reported
trillion in 1979; the ratio of 2005 to 1979 is 4.599. The inflation
damages in current-year dollars; Iy = inflation adjustment;
multiplier for 1979 was 2.275, so the inflation-corrected wealth
RWPHUy = real wealth per housing unit adjustment; and
adjustment for 1979= 4.599/ 2.275= 2.021, exactly as in PL05.
HU2005/y = coastal county housing unit adjustment.
The number of U.S. housing units in 1979 is estimated to be
As an example, here is how damage from Hurricane Frederic
86,438,040 共based on a linear interpolation of 68,679,030 units in
is calculated: Dy = $ 2 , 300, 000, 000; Iy = 2.275; RWPHUy = 1.424;
1970 and 88,411,263 in 1980兲. U.S. housing units in 2005 were
and HU2005/y = 1.549, or $2 , 300, 000, 000共⫻兲2.275共⫻兲1.423
estimated to be 122,725,123 共using a linear extrapolation from
共⫻兲1.549= $ 11, 537, 923, 783 共this is the actual normalized dam-
1990–2000兲. The U.S housing unit multiplier is thus the ratio of
age result for Frederic calculated using nonrounded multipliers兲.
the 2005 estimate to the 1979 estimate, or 1.420. Thus, the final
Frederic caused $2.3 billion in total damage when it made
wealth multiplier for 1979 is the real wealth multiplier of 2.021
landfall in 1979. If this same storm were to have occurred in
divided by the U.S. housing unit multiplier of 1.420, which equals
2005, it would have caused an estimated $11.5 billion in total
1.424. Therefore each housing unit in the United States contains
damage under the CL05 approach to normalization.
共on average兲 1.424 times more wealth in 2005 than did each hous-
ing unit in 1979.
The final multiplier in CL05 is county housing units, and as Discussion of Results of Normalization
with other U.S. Census information, housing unit data are pro-
vided by decade, and linear interpolation 共extrapolation兲 provides Fig. 3 shows U.S. hurricane damages from 1900–2005 adjusted
the data for all years 1940–2005. Specifically, Joel Gratz updated only for inflation, showing a clear increase in losses. The dark
a spreadsheet of housing unit data compiled by D. Collins for line represents an 11-year centered moving average. Figs. 4共a–c兲
Collins and Lowe 共2001兲 based on U.S. Census 共2006兲. At the show the summarized and individual results for the two different
time of our research the census only had this information by approaches to normalization for the complete data set. The results
county by decade in nondigital media 共Bureau of the Census of PL05 and CL05 tend to be very similar, with larger differences
1990兲. Housing units for 1900–1939 were estimated based on further back in time.
extrapolating back in time the county-level relationship of popu- Further details can be seen in the tables. Table 2 shows the top
lation and housing units from 1940–2005. 50 damaging events, ranked by PL05, along with the correspond-
From the county-level housing unit data, a housing unit mul- ing ranking of CL05. Under both approaches, the 1926 Great
tiplier was calculated based on the ratio of county housing units in Miami hurricane is estimated to result in the largest losses at $140
2005 to that of the year in which the storm originally made land- billion–$157 billion. Hurricane Katrina is second under both
fall. For example, the 1979 storm Frederic affected Baldwin and normalization schemes. The years 2004 and 2005 stand out as
Mobile counties in Alabama and Jackson County in Mississippi. particularly extreme, with 7 of the top 30 most damaging 共nor-
The sum of the housing units for these counties in 2005 is malized兲 storms over 106 years. No other 2-year period has more
312,749 compared to 201,946 in 1979. Thus the population ad- than 3 top 30 storms 共1944–1945兲. Of particular note is the rapid
justment for the 1979 storm Frederic is 312, 749/ 201, 946 increase in estimated damage for historical storms as compared to
= 1.549. Pielke and Landsea, who, for instance, estimated that the 1926
Category of storm Mean damage 共$ million兲 ⬍1 million people 1–3 million people ⬎3 million people
PL05 average damage 共$ million兲 by 2005 population value
Tropical/subtropical 140 170 共45兲 90 共7兲 1,930 共7兲
1 1,200 340 共35兲 1,400 共5兲 6,030 共6兲
2 2,200 1,400 共21兲 3,000 共9兲 4,300 共6兲
3 7,000 5,800 共38兲 5,600 共11兲 13,800 共9兲
4 30,000 11,900 共8兲 18,000 共2兲 63,600 共5兲
5 26,400 11,700 共2兲 55,800 共1兲 — 共0兲
CL05 average damage 共$ million兲 by 2005 population value
Tropical/subtropical 140 180 共45兲 90 共7兲 1,800 共7兲
1 1,300 400 共35兲 1,200 共5兲 6,300 共6兲
2 2,200 1,500 共21兲 2,800 共9兲 4,000 共6兲
3 7,000 6,000 共38兲 5,800 共11兲 13,000 共9兲
4 28,500 13,300 共8兲 17,200 共2兲 57,100 共5兲
5 26,500 13,500 共2兲 52,300 共1兲 — 共0兲
Note: Although only 14 major hurricanes have made landfall in an area with greater than 1 million people, this table illustrates the pronounced increase
in vulnerability from a larger population. The average damage of a Category Four hurricane increases 3.5 times when making landfall in an area with ⬎3
million people compared to 1–3 million people 共parentheses denote number of storms in that cell兲.
conditions that is not uniform between events. For example, de- today would lead to significant shortages in the affected areas and
mand surge will be greater in periods of strong economic activity result in inflationary pressures. Thus, our historical estimates may
and low unemployment due to the lack of slack resources. Local be considered conservative.
economic conditions will also have an effect, as will the proxim- Another important factor is mitigation and the implementation
ity of losses in time and space 共the demand surge in the 2004 of stronger building codes. There is considerable evidence that
Florida hurricanes was greater than would have been the case had strong building codes can significantly reduce losses; for ex-
the four major loss events occurred in different years兲.
ample, data presented to the Florida Legislature during a debate
The normalization methodologies used in this paper assume
over building codes in 2001 indicated that strong codes could
that demand surge is uniform over time. To the degree that past
losses were relatively smaller in the context of the economy of the reduce losses by over 40% 共IntraRisk 2002兲. As strong codes have
time than they would be today, the methodology may understate only been implemented in recent years 共and in some cases vary
the size of the loss in current dollars and vice versa. A good significantly on a county-by-county basis兲, their effect on overall
example of this might be the Miami hurricane of 1926, which was losses is unlikely to be large, but in future years efforts to im-
a smaller proportion of the national economy than a similar event prove building practices and encourage retrofit of existing struc-
would be in 2005. Certainly, an event larger than $100 billion tures could have a large impact on losses.
Appendix